Engineering detection intelligent equipment and system
By designing intelligent equipment for engineering inspection and using deep learning models for image recognition and data analysis, the problems of low efficiency and error-proneness in the existing technology have been solved, and more efficient and accurate engineering quality and safety inspections have been achieved.
Patent Information
- Application Number
- CN202411372972.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-29
- Publication Date
- 2025-01-21
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing engineering quality and safety inspection methods are inefficient and prone to errors, and human factors make it difficult to guarantee fairness and rationality.
Design an engineering inspection intelligent device, including a data acquisition module, a processing module and an interaction module. It uses a deep learning model for image recognition and data analysis, establishes an evaluation system and outputs inspection reports and inspection scores.
Improve the accuracy and efficiency of engineering detection, reduce artificial intervention, and ensure the fairness and rationality of the test results.
Smart Images
Figure CN119339132A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of engineering quality and safety detection, and in particular to an engineering detection intelligent device and system. Background Art
[0002] With the development of science and technology, artificial intelligence technology has been widely used in various fields. In the field of engineering inspection, traditional inspection methods often require manual operation, which is inefficient and prone to errors. Therefore, how to use AI technology to improve the accuracy and efficiency of engineering inspection has become an urgent problem to be solved.
[0003] In the early stages of existing engineering quality and safety testing, evaluation systems and standards need to be manually configured offline according to the needs of each customer. There are many types of systems for each customer, and it is complicated and tedious to configure hundreds of customers at the same time, and the workload is large.
[0004] Secondly, the existing project quality and safety inspection relies on on-site observation or testing by professionals, who use their professional experience and knowledge to conduct manual inspections of the on-site construction quality, visual quality, and construction machinery safety. The workload is huge, the number of personnel involved is large, the human control factors are large, and fairness and rationality are difficult to guarantee.
[0005] Finally, after the on-site professionals have completed the inspection, you will need to organize the photos and reports yourself, which takes a long time. Summary of the invention
[0006] In order to solve one or more problems existing in the prior art, the present invention provides an intelligent device for engineering detection. The technical solution adopted by the present invention to solve the above problems is: an intelligent device for engineering detection, the device comprises: a data acquisition module, a processing module and an interaction module, the data acquisition module is used to collect image data to be detected, the processing module is used to perform data calculation, and the interaction module is used for data transmission and human-computer interaction.
[0007] and a system applied on the device, the system comprising: an identification module, an evaluation system establishment module and a data annotation module;
[0008] The recognition module includes: a deep learning model, through which the image data to be detected is automatically recognized and corresponding image recognition data is obtained;
[0009] The data annotation module annotates the image recognition data and forms label data corresponding to the image recognition data, wherein the label data includes a building part, a damage degree and a damage type;
[0010] The engineering quality and safety inspection system table and delivery system table acquired by the system perform automatic vocabulary recognition through a deep learning model, and the automatic vocabulary recognition obtains system recognition data by splitting and recognizing the vocabulary, and the system recognition data includes damage problems, implementation standards and solutions corresponding to various building parts;
[0011] The evaluation system establishment module aggregates the system identification data into different customer systems, establishes a mapping relationship between the system identification data and the label data, establishes an evaluation system corresponding to the customer system, and the interaction module outputs the data of the evaluation system to the user end;
[0012] The mapping relationship is the correspondence relationship and the hierarchical relationship between the system feature words in the customer system and the annotation words in the label data.
[0013] In some embodiments, the evaluation system establishment module aggregates the modification data input by the user end and corresponds it to the customer system.
[0014] In some embodiments, the evaluation system establishment module establishes the mapping relationship through the deep learning model.
[0015] Furthermore, the vector k1 of the system feature word and the vector k2 of the annotation word are defined by the deep learning model, and the reliability of the mapping relationship is p, then Where n is the number of dimensions of the vector, k1i is the component of k1 in the i-th dimension, k2i is the component of k2 in the i-th dimension, e is a natural constant, q is the total vocabulary, and λ is the deviation coefficient, λ=1.2×10 ― 5 ; The reliability p is used to determine whether the mapping relationship is credible;
[0016] The data acquisition module acquires the detection angle w1 of the image data to be detected, the system feature word is set with a standard angle w2, w1∈[0,180], w2∈[0,180], the correction coefficient of the reliability p is f, m is the empirical constant, m=9.437, and the corrected reliability is p1, p1=p*f.
[0017] In some embodiments, the evaluation system is provided with scoring criteria corresponding to the engineering quality and safety detection system table and the delivery system table, and the calculation parameters of the scoring criteria include the system identification data;
[0018] The label data determines the corresponding system identification data and the customer system through the mapping relationship, and then calculates the detection score of the image data to be detected through the scoring standard.
[0019] In some embodiments, the evaluation system organizes the data of the customer system into a test report according to a preset template.
[0020] In some embodiments, the user performs a keyword search on the tag data, the system identification data, the customer system, and the evaluation system through the interaction module.
[0021] In some embodiments, the user performs selection association on the tag data, the system identification data, the customer system and the evaluation system through the interaction module.
[0022] The technical effects achieved by the present invention are: by adopting a trained deep learning model to detect the acquired data, its accuracy and efficiency are higher than the traditional manual inspection and scoring method; through the above-mentioned evaluation system establishment module, a more comprehensive analysis of the data can be performed, and the analysis requirements can be adjusted according to actual needs, and finally test reports and test scores in different formats and requirements can be output for user reference; through the above-mentioned interactive module, users can adjust and set the system and provide an intuitive user-side interface; the detection efficiency and quality of engineering inspection in image data can be improved. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] Figure 1 is a schematic block diagram of a device of the present invention;
[0024] Figure 2 A schematic block diagram of the system of the present invention. DETAILED DESCRIPTION
[0025] In order to make the above-mentioned objects, features and advantages of the present invention more understandable, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from this description, and those skilled in the art can make similar improvements without violating the connotation of the present invention, so the present invention is not limited by the specific embodiments disclosed below.
[0026] like Figure 1 As shown, the present invention discloses an intelligent engineering detection device, which includes: a data acquisition module, a processing module and an interaction module, the data acquisition module is used to collect image data to be detected, the data acquisition module collects detection angles corresponding to the image data to be detected, the processing module is used to perform data calculation, and the interaction module is used for data transmission and human-computer interaction.
[0027] and a system applied on said device, in combination Figure 2 As shown, the system includes: an identification module, an evaluation system establishment module and a data annotation module;
[0028] The recognition module includes: a deep learning model, through which the image data to be detected is automatically recognized and corresponding image recognition data is obtained;
[0029] The deep learning model can be a trained CNN model, and the training image data is annotated building crack photos, building leakage photos, control group photos, etc., to achieve image recognition function. Similarly, by providing vocabulary template files for training, vocabulary recognition, splitting, and fusion functions can be achieved;
[0030] The data annotation module annotates the image recognition data and forms label data corresponding to the image recognition data, wherein the label data includes a building part, a damage degree and a damage type;
[0031] The engineering quality and safety inspection system table and the delivery system table acquired by the system are automatically recognized by the deep learning model, and the automatic recognition of vocabulary is performed by splitting and recognizing vocabulary to obtain system recognition data, and the system recognition data includes damage problems, implementation standards and solutions corresponding to various building parts;
[0032] The evaluation system establishment module aggregates the system identification data into different customer systems, establishes a mapping relationship between the system identification data and the label data, establishes an evaluation system corresponding to the customer system, and the interaction module outputs the data of the evaluation system to the user end;
[0033] The evaluation system establishment module aggregates the modified data input by the user end and corresponds it to the customer system, thereby realizing automatic modification and adjustment of the system identification data.
[0034] Specifically, the evaluation system establishment module establishes the mapping relationship through the deep learning model, and the mapping relationship is the correspondence and the upper and lower relationship between the system feature words in the customer system and the annotation words in the label data. Through the mapping relationship, the label data obtained from the image data to be detected can be quickly mapped to the system identification data in the customer system, thereby facilitating the evaluation system to sort and score the system identification data corresponding to the image data to be detected.
[0035] Because different customers have different inspection system tables and delivery system tables, the key scoring criteria involved are essentially close to unified standards (national industry standards). The difference lies in the division and refinement of major categories, and the detailed division and scoring criteria of each project under the major categories. The major categories, projects and standards are different in textual expression. This difference is mainly due to the hierarchical and similar relationships in the name definitions. For example, the acceptance project concrete structure engineering in the first delivery system is building structure engineering in the second delivery system. These are similar or substantially the same acceptance projects.
[0036] The deep learning model is used to learn and train the existing form templates and form the corresponding mapping relationship, so as to quickly convert the recognition results of the image data to be detected into parameter items of different detection system tables and delivery system tables, thereby greatly reducing the need for manual form filling and proofreading, reducing labor costs, and improving acceptance and inspection quality.
[0037] It should be pointed out that the label data is formed by annotating the internal database of the deep learning model, and the annotated words therein are internally defined standard words. Then, based on the annotated words, the internal standards are mapped to different external detection system tables and delivery system tables, thereby realizing the corresponding conversion of internal and external vocabulary and sentences.
[0038] The training of the deep learning model includes: data preparation, data preprocessing, manual annotation and definition of data, data feature extraction and segmentation, model algorithm determination, model training, model parameter adjustment and optimization, model recognition, model result processing feedback and model deployment, etc. The deep learning model can be an existing CNN model, or other models that provide image recognition and text recognition functions.
[0039] Among them, in establishing the mapping relationship, the mapping relationship between the system feature words in the customer system and the annotation words in the label data can be that the deep learning model realizes the hyponym mapping and the same mapping based on the synonym word set in the dictionary resource.
[0040] Specifically, in order to determine whether the mapping relationship is credible, a reliability p is introduced into the mapping relationship. The vector k1 of the system feature word and the vector k2 of the annotation word are defined by the deep learning model. Among them, n is the number of dimensions of the vector, k1i is the component of k1 in the i-th dimension, k2i is the component of k2 in the i-th dimension, e is a natural constant, q is the total vocabulary, λ is the deviation coefficient, and the system determines the credibility according to the size of the reliability p. Accordingly, according to the actual situation, the system sets different trust thresholds corresponding to the reliability p, and determines whether to adopt the current mapping relationship by comparing the sizes.
[0041] Furthermore, the size of the deviation coefficient λ is related to the training time, overall accuracy and total vocabulary of the deep learning model, which can be λ=1.2×10 ― 5 , which is the main tuning parameter. Among them, is the similarity calculation, e ― λ q Calculate reliability.
[0042] Since operators often take wrong shots, multiple shots, or blur when actually taking the image data to be detected, the final recognition accuracy is reduced. Multiple shots and blur can be filtered out, but filtering out wrong shots caused by not following the shooting instructions is relatively difficult and troublesome. Not following the shooting instructions means that the operator does not place the image of the building to be detected at the center of the guide circle, and there is a problem with the shooting angle. The problem with the shooting angle can easily lead to errors in the recognition of building parts in image recognition.
[0043] Specifically, the data acquisition module includes a camera, a gyroscope and an accelerometer, and the image data to be detected is formed by shooting with a camera. The detection angle is the shooting pitch angle at the time of shooting, wherein the camera is shooting directly at the ground at 0 degrees and shooting directly at the ceiling at 180 degrees. The parameters obtained by the gyroscope and the accelerometer can be calculated by the processing module to obtain the corresponding detection angle. The interactive module may include a touch screen, a wired or wireless data transmission module.
[0044] The data acquisition module acquires the detection angle w1 of the image data to be detected, the system feature word is set with a standard angle w2, w1∈[0,180], w2∈[0,180], the correction coefficient of the reliability p is f, m is an empirical constant, which can be m=9.437. The corrected reliability is p1, p1=p*f. The size of m is related to the error of w1 and the accuracy of recognition, and it is a major parameter for parameter adjustment. For example: the positive angle of the structural column (indoor), w1=150 degrees. The corrected reliability p1 can further exclude images taken by operators not following the instructions, thereby ensuring the accuracy of later judgments.
[0045] It should be pointed out that the measurement of the detection angle requires the operator to perform angle correction and reference plane determination on the client side of the shooting before accurate measurement can be performed.
[0046] Specifically, the evaluation system is provided with scoring criteria corresponding to the engineering quality and safety inspection system table and the delivery system table, and the calculation parameters of the scoring criteria include the system identification data;
[0047] The label data determines the corresponding system identification data and the customer system through the mapping relationship, and then calculates the detection score of the image data to be detected through the scoring standard;
[0048] The evaluation system organizes the data of the customer system into a test report according to a preset template, and the preset template corresponds to the test report template in the engineering quality and safety test system table and the delivery system table;
[0049] The user selects and associates the label data, the system identification data, the customer system and the evaluation system through the interactive module. Since some words and sentences in the engineering quality and safety inspection system table and the delivery system table are not similar or do not belong to the upper and lower relationship, but there is a reference, guidance and progressive relationship, the user needs to manually select and associate to ensure that the inspection report and inspection score can be output normally. For example: the annotation words about the damage type in the label data will point to the solution and implementation standard in the system identification data.
[0050] Specifically, the user performs a keyword search on the tag data, the system identification data, the customer system and the evaluation system through the interactive module, so that the user can manually adjust and view the data.
[0051] In summary, by using a trained deep learning model to detect the acquired data, its accuracy and efficiency are higher than the traditional manual inspection and scoring method; the above-mentioned evaluation system establishment module can realize a more comprehensive analysis of the data, and adjust the analysis requirements according to actual needs, and finally output inspection reports and inspection scores in different formats and requirements for user reference; the above-mentioned interactive module enables users to adjust and set the system, and provides an intuitive user-side interface; it can improve the inspection efficiency and quality of engineering inspection in image data.
[0052] The above-described embodiments only express one or more implementation modes of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the patent of the present invention. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the attached claims.
Claims
1. An intelligent engineering detection device, characterized in that: The device comprises: a data acquisition module, a processing module and an interaction module, wherein the data acquisition module is used to acquire image data to be detected, the processing module is used to perform data calculation, and the interaction module is used for data transmission and human-computer interaction.
2. A system, comprising the device according to claim 1, characterized in that: The system includes: an identification module, an evaluation system establishment module and a data annotation module; The recognition module includes: a deep learning model, through which the image data to be detected is automatically recognized and corresponding image recognition data is obtained; The data annotation module annotates the image recognition data and forms label data corresponding to the image recognition data, wherein the label data includes a building part, a damage degree and a damage type; The engineering quality and safety inspection system table and the delivery system table acquired by the system are automatically recognized by the deep learning model, and the automatic recognition of vocabulary is performed by splitting and recognizing vocabulary to obtain system recognition data, and the system recognition data includes damage problems, implementation standards and solutions corresponding to various building parts; The evaluation system establishment module aggregates the system identification data into different customer systems, establishes a mapping relationship between the system identification data and the label data, establishes an evaluation system corresponding to the customer system, and the interaction module outputs the data of the evaluation system to the user end; The mapping relationship is the correspondence relationship and the hierarchical relationship between the system feature words in the customer system and the annotation words in the label data.
3. The system according to claim 2, characterized in that The evaluation system establishment module aggregates the modified data input by the user end and corresponds it to the customer system.
4. The system according to claim 2, characterized in that The evaluation system establishment module establishes the mapping relationship through the deep learning model.
5. The system according to claim 4, characterized in that The vector k1 of the system feature word and the vector k2 of the annotation word are defined by the deep learning model, and the reliability of the mapping relationship is p. Wherein, n is the number of dimensions of the vector, k1i is the component of k1 in the i-th dimension, k2i is the component of k2 in the i-th dimension, e is a natural constant, q is the total vocabulary, and λ is the deviation coefficient; the reliability p is used to determine whether the mapping relationship is credible.
6. The system according to claim 5, characterized in that The data acquisition module acquires the detection angle w1 of the image data to be detected, the system feature word is set with a standard angle w2, w1∈[0,180], w2∈[0,180], the correction coefficient of the reliability p is f, m is an empirical constant, and the corrected reliability is p1, p1=p*f.
7. The system according to claim 2, characterized in that The evaluation system is provided with scoring criteria corresponding to the engineering quality and safety inspection system table and the delivery system table, and the calculation parameters of the scoring criteria include the system identification data; The label data determines the corresponding system identification data and the customer system through the mapping relationship, and then calculates the detection score of the image data to be detected through the scoring standard.
8. The system according to claim 2, characterized in that The evaluation system organizes the data of the customer system into a test report according to a preset template.
9. The system according to claim 2, characterized in that The user performs a keyword search on the tag data, the system identification data, the customer system and the evaluation system through the interactive module.
10. The system according to claim 2, characterized in that The user selects and associates the tag data, the system identification data, the customer system, and the evaluation system through the interaction module.